# Dashboard Overview The Karmaflow.ai dashboard is the main workspace overview for your team. It opens on **Now** — the hour around this moment, on a map — and keeps the date-ranged analytics on the tabs beside it: **Overview** (one page: channels, days, hours, start pages, where conversations fail), **Conversation Insights** (the cards your workspace configures), then your own tabs. ## Now — the hour around this moment The first tab is a sixty-minute window centred on now. The last thirty minutes are what happened: every chat, call, text, email, form submission and agent run, each drawn where it happened. The next thirty are what is about to happen: workflows due to run, campaign sends queued, send-later emails and texts, tasks due, SLA deadlines, held AI actions that will run unless someone declines them, the dial queue, and jobs starting in a connected field-service system. - **The headline** names where the hour is happening — one city, one state or province, one country, or *The world* — and the hero number beside it is the count of conversations in the last thirty minutes, with how many places they span and whether this hour is quieter or busier than your workspace's own norm for this weekday and hour (*Learning your rhythm* until a few weeks of history exist). A chat counts once the visitor has written something, from their first message: a visitor who opened the chat widget and has not typed yet is not a conversation, and appears as its own line under the number instead (*2 with the chat open, not written yet*). A queue of calls waiting to be dialled is said as such (*51 calls queued to dial*), not as something scheduled. The sentence under the headline is written from the counts on the screen: how many live, how many waiting on a person (a live chat whose agent asked for a person and nobody has picked it up yet counts here), how many of the conversations that finished were handled without needing a person (*11 of 12 finished without needing a person*), what is scheduled. That last figure is cautious on purpose: a conversation where the agent asked for a person, whatever happened next, a transferred call, an escalation, or a call a person made themselves is never counted as handled alone, and a failed call (no answer, busy) is not counted at all. - **Colour by** switches what the lights mean: **Outcome**, **Channel** or **Sentiment** in one click, **Status**, **Agent** or **Campaign** under **More…**. The legend on the same row carries a count per class, and the **Unplaced** count beside it is always visible — nothing is silently dropped from the map. Click or tap **Unplaced** to see what is not on the map and why: a chat whose visitor's connection could not be looked up, an email whose contact's place is not known, and so on. An agent run that happened inside a chat or call is drawn where that conversation is, and its card links to it. - **The stage** opens where your conversations are. A workspace with a pinned place lands on it (pinning is you telling the map where home is); otherwise the hour's footprint sets the altitude — one city lands over the city, one state or province over the region, one country over the country, and a footprint across continents opens on the planet turned to face your conversations. With street precision on and exact pins near home, the camera lands one step closer so the first frame already shows pins on streets. Pale at the planet, full colour in the streets; your busiest cities are named and your pinned places ringed and labelled. The pills under the map (**World** and your places) and the **− / +** control move the camera anywhere, and where you leave it is where it reopens for the rest of your visit. The words in the corner say where it is (*Planet view*, *City view · Denver HQ*). When the **Weather** overlay is on, a tile in the corner shows the temperature and conditions for the place in **In focus** at any altitude (or, with *Everywhere* in focus, for wherever the camera is once you zoom in). The base map is deliberately quiet — pale land, no shops, railways or airports, street names only when you are close — so the lights are what you see. - **The map needs WebGL.** If your browser has graphics acceleration switched off, the stage says so and shows how to turn it back on (**Use graphics acceleration when available** in the browser's system settings, then restart and reload); the headline, the **In focus** card, the moments and the bands below keep working without it. - **Every light is placed only as precisely as the platform truly knows.** A city-level fact (a caller's registered city, a visitor's connection) is a soft disc the size of the city; an exact location is a pin only when your workspace has opted in to street precision under **Settings → Access & security → Location & Map Precision**. Otherwise the map stays at city scale for everyone, and the card's footnote says so. - **Age is drawn as fade**: a new light pulses for two minutes, then dims across the half hour. Live conversations stay bright. - **In focus** — the card over the stage — is about one place at a time: your first pinned place, or the busiest city of the hour, or everywhere (pick from its menu). It shows that place's conversations in the half hour with a five-minute sparkline and a bar split by the chosen colour, the latest moments there with their agent and outcome (each a link to its record) — ordered by what needs you first: conversations waiting on a person, then live conversations with a known contact, live anonymous ones, then those that have ended (known contacts before anonymous visitors), newest first within each. A moment with no title and no known contact reads *Website visitor* or *Caller*. Under a live chat that has no title yet, the visitor's last words appear in quotes, for people who may read chats, and the age reads *last message 3 min ago*. Where the workspace has turned on AI topics, a short AI-written line saying what the visitor wants (*Price of a home battery*) takes the quotes' place, marked with a sparkle; the record shows both; a chat waiting on a person reads *waiting 6 min*. Choosing a place in the card's menu moves the map there; **Everywhere** turns it back to the world. A chat waiting on a person links straight to the operator panel; what is coming up, and one action — **Explore Denver HQ** flies the camera in; **Return to the world** brings it back. Click a light on the map or a moment in the card to read that record instead; selecting a queued campaign send lights up its recipients on the map. - **Beneath the stage**, one quiet row: ticks for the overlays your workspace has switched on (traffic, weather, air quality, daylight, public holidays) — overlays that are switched off for the workspace show greyed as *off* and link to where a tenant admin turns them on (**Settings → Workspace → Map Layers & Places → Overlays**) — and for each of your layers, with their freshness in the tooltip; then the slider across the hour (back to see what had happened by a given minute, forward to see what will have fired; press play to follow the clock again), **Yesterday**, **Chime**, **Put this on a screen** with a link button beside it, and the live clock with your precision level. - **Yesterday in sixty seconds.** **Yesterday** replays the previous day on the map in about a minute: the slider runs from midnight to midnight, each light appears where and when it happened and fades over the hour, and the headline is written for the hour the cursor is in ("10:00 — 31 conversations, mostly calls, mostly in Ottawa"; quiet stretches are folded). Pause, drag the slider, restart, or go **Back to now**. - **Put this on a screen** (Wall mode) turns the tab into a full-screen board: the stage fills the screen, the headline is large, a ticker of the latest moments runs under it, and the places your workspace has pinned are cycled every twenty seconds. It is meant to be read from across a room: there is no greeting by name, no controls apart from **Leave wall mode** and the clock, and no visitor's words on any row. Leave it with **Esc** or **Leave wall mode**. It is the same page at `?wall=1`, so a dispatcher's screen or a lobby display can open it directly: the link button beside **Put this on a screen** copies that address, for a TV or a second monitor signed in to the workspace. - **Chime.** Tick **Chime** and the tab plays a soft two-note sound whenever someone starts waiting on a person or a new card lands in **Needs you**. It is off until you turn it on, and the choice belongs to this browser only. Browsers play sound only after you have used the page, so turning it on plays it once, and after a reload the tab may stay quiet until your first click on it. - **AI topics on live chats** are a workspace choice, off by default. A tenant admin sees *Show an AI topic on each live chat?* under Latest moments, with **Turn on**. Once on, each live chat gets one short line written by AI from the conversation so far, until the chat is titled when it ends. It is only shown to people who may read chats, never on a screen view, and never includes a name, an email address or a number. It runs only while someone who may read chats has the tab open, for at most five chats a minute, and it also needs Karmaflow to enable the feature for your workspace: until then the card says it is waiting for that. - **Since you last looked.** When you come back to the tab after a few minutes away, a line under the headline says what happened meanwhile (*Since you looked 24 min ago: 2 new conversations · 1 ended · 1 waiting on a person.*), and the new moments carry a dot in Latest moments. **Got it** clears it. It is remembered per browser; an absence longer than the half hour on screen says so. - **The browser tab's title** reads *(2) Needs you · Karmaflow.ai* while something in Needs you is waiting on you, so a tab left in the background still tells you. Cards you hid do not count. Below the map are five personal bands: **Needs you** (chats waiting for a person, campaign questions, held approvals, breached SLAs, failed agent runs — and, for tenant admins, the Settings hub's own attention items), **Your places** (the pages you open most, tilted toward what you open at this hour of the week), **Pick up where you left off** (records you viewed recently), **Today so far** (conversations started today on each channel, live ones included, counted the same way the map counts them: a chat counts once the visitor has written. Each tile says how many are live now and how many usually came in by this time of day, from the same weekday over the last four weeks. A channel your workspace has not used in four weeks is left out. **Full analytics** counts finished conversations, so its figure for today can be lower) and **What changed** (recent changes in the workspace). - When nothing is waiting, the band says so: *Needs you · Nothing is waiting on you right now.* - **Hiding a Needs you card.** The **×** on a card hides it for you alone; it changes nothing for anyone else and does not resolve what the card is about. A hidden card comes back by itself when it gets worse (more items, a higher severity, or something new since you hid it) or after a day for a red card and a week for an amber one. The band heading always says how many you have hidden; **Show** lists them with **Unhide**. - **Security findings** reach Needs you for tenant admins when a high or critical finding is open. To quiet one for the whole workspace, open **Review security findings** and choose **Mark as seen** on the finding: it stops asking for attention on the dashboard and the settings hub until it becomes more severe, and the finding shows who marked it and when. Karmaflow.ai still reviews it; its status does not change. - **What changed** keeps a settings or agent change as its own line, with who made it. Changes to records are grouped by kind, action and who made them, within the same hour — *3 contacts created · An agent* — and a group opens on click to the records, each a link. When the band has room, groups are shown open. A bulk action reads as one row with its count (*4 contacts deleted*); repeated saves of the same setting by one person read as one row (*… · 3 edits*); records created as a side effect, such as activity entries, are left out. A change made through an agent's tool reads *An agent*; one made by a person shows their name. Your scope follows your role: tenant admins, super admins and anyone with the analytics-manage permission see the whole workspace; everyone else sees the interactions linked to contacts they own. ## Overview — how much, and how it is moving The **Overview** tab answers one question at a glance: how much is coming in, and how it moves day by day. The score cards give the totals; the chart under them gives the days. Every figure counts the same **analyzed conversations** (closed, billable, from your production agents) that [Conversation Insights](/help/reports/conversation-insights) and your own analytics tabs count, so a figure here is the figure there. *When in the week*, *Where do chats start?* and the Topic map lead [Conversation Insights](/help/reports/conversation-insights) since October 2026. - **The AI briefing** sits at the top: a timezone-aware greeting, a short specific summary of how things are going (conversation sentiment and outcomes, channel volume, CRM movement), tone-coloured insight chips and suggested next steps. It runs on its own fixed window and counts *interactions* — chats, calls, texts and emails sent — where the cards below count conversations; the chip says both. It is cached for the day; **Refresh** regenerates it. Admins, managers and anyone with the analytics-manage permission see it across every account; everyone else sees it scoped to the accounts they own, and the card says when it is. - **The score cards** are the page's headline: **Conversations** (with how many were analyzed), one card per channel that had any (**Chats**, **SMS conversations**, **Calls**, **Email threads**), **Emails sent** — by every sender: your mail strategies and your chat, voice and task agents (follow-ups, escalations, campaign sends), production agents only — and **Resolved**, the share of analyzed conversations that ended resolved or likely resolved. An email sent belongs to a chat, a call or a campaign, so it is never counted as a conversation; a channel with none in range has no card and no chip. Pick **Email** on a workspace whose customers never write in by email and the page says there are no email conversations — an email conversation is a thread a customer started — and still charts the emails your agents sent, with **Show every channel** beside it. Click a channel's card, or the chips under the cards, to narrow the page to it; the emails sent leave when you narrow to a channel other than email or to who handled the conversation (a send has no "handled by"). With **Compare** on (*vs previous period*, *vs same period last year*) every card shows its change — a percent for a count, points for the resolved share — the daily chart adds a dashed line of the previous window (with Compare on, a channel that had none in the previous window reads *New · none in the previous period*, and a share that did not move reads *No change*), and a **What changed** box lists the biggest movers. Your comparison choice is remembered on this browser. - **How is volume moving day by day?** — one line per channel, each read against the same axis from zero (lines, not stacked bands, so a small channel never sits on top of a large one and reads larger), plus an **Emails sent** line beside them. The emails line is not part of the conversations' total: the sentence names the leading channel and its peak, then the emails sent; hovering a day shows each line and the conversations' total; today, while it is still under way, is drawn dashed and marked *today so far* rather than read as a drop; clicking a channel's point lists that day's conversations (the emails line has no list here — **Agents → Reports & Logs → Sent Emails** has every send). - **Agent performance** (quality scores across every production agent, **See all agents →**) and **What's New** close the page. Every card has the same controls as a widget on a tab of your own: hover a mark for its figure, click it to list the conversations behind it and open one, switch to the **table** view, and **⋯ → Export CSV / Export PNG**. The tab itself is read-only — it is the same for everyone — but **⋯ → Duplicate to my tabs** copies it into a tab of your own, where you can add, edit, rearrange and remove cards (see [Build your own analytics tabs](/help/reports/custom-analytics-tabs)). **Where the old tabs went.** *Channels & Activity* and *Chat Pages* folded into this page on 2026-10-21: the channel cards, the daily chart, the busiest hours and the chat-start pages are the cards above. An old link to either tab opens Overview and says so. ## What's New The **What's New** card surfaces the latest release notes right on the dashboard. When a new version ships, the card re-opens itself so you don't miss it; minimize it once you've caught up and it stays out of the way until the next release. The same notes live in the [What's New](/help/whats-new/overview) help article. ## Filtering The date range picker at the top of the page sets the window for Overview, Conversation Insights and your own tabs alike (the AI briefing keeps its own). On Overview and Conversation Insights, the line above the cards and the chips under it narrow every card by channel and by who handled the conversation — your AI agents, your team, or both.